Creating accurate precipitation maps using machine learning
Using machine learning, researchers from the University of Hawaii and the East-West Center improved the accuracy of monthly precipitation maps. Using this method, scientists detected errors in existing maps. They obtained systematized data on monthly precipitation in Hawaii for the past 30 years, from 1990 to 2019. They used readings from more than 600 weather observation stations across the islands to compile them.
The experiment showed how a machine learning algorithm can evaluate and classify unrealistic images (in this case, precipitation maps). Researchers are developing methods to automate the creation of daily precipitation maps and other systematized data products for the state of Hawaii.
Such maps will enable the development of a statewide decision-making system that addresses fire, flood, and drought risks. High-quality, real-time rainfall maps will allow less time to process the data.
‘Machine learning’ improves rainfall mapping, water plans
Technologies: Artificial intelligence.
Geography: US, Hawaii
Developer: University of Hawaii, East-West Center